Expert witnesses in civil proceedings are legally required to be independent. The current recusal mechanism relies on self-declaration and adversarial challenge — a slow, qualitative process with no quantitative standard for assessing the network proximity between an expert and the parties they serve.
This paper applies network proximity analysis — using Erdős-number methodology on academic co-authorship and supervisory graphs — to demonstrate that the existing procedural standard is structurally inadequate. In the case study examined, the Azevedo-Henriques relationship places the expert witness at Erdős distance 1 from the commissioning party: a level of proximity held by approximately 500 people globally, indistinguishable from direct collaboration, and automatically within the statutory appearance standard for recusal — yet one that went unchallenged under existing procedure.
The paper proposes a formal network proximity threshold — grounded in scale-free network statistics — as a required component of expert appointment procedure across European civil proceedings.
This paper has been formally received by the Centro de Estudos Judiciários (CEJ), entered into the permanent collection of the Biblioteca Armando Leandro, and will be distributed to its teaching faculty. The CEJ is the institution responsible for the professional training of all judges in Portugal.
The Network Proximity Check at kyc.co/network-proximity operationalises the methodology of this paper. It computes the Network Proximity Score (NPS, 0–100) from academic co-authorship data, supervisory relationships, and institutional affiliations — producing a quantitative proximity assessment before an expert is appointed, not after they have already testified.
Carroll, S. (2026). Network Proximity and Expert Impartiality: An Empirical Analysis of Academic Network Proximity as a Structural Test for Expert Recusal in European Civil Proceedings. Working paper. SSRN 6905898. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6905898
Legal proceedings treat qualification as a proxy for expertise. The two are not the same thing. A credential certifies that the holder met an educational standard at a past point in time. Expertise requires something different: demonstrated performance in a specific technical domain at a sufficient depth and recency to be relevant to the question at issue.
The gap between formal qualification and actual domain competence — Competence Distance — is not currently a formal object of procedural scrutiny. Courts do not have a methodology for measuring it. The paper identifies the signals that make this gap detectable before testimony: publication distance from the specific technical domain, recency of domain-specific engagement, the ratio of generalist to specialist output in the expert's documented work, and the structure of their prior instruction record.
These signals are operationalised as the Specialisation Distance Differential (SDD) — a scored composite (0–100) that can be computed from publicly available professional data and applied as a pre-appointment screening instrument. A high SDD does not disqualify an expert; it raises a procedural question that the appointment process is currently not equipped to ask.
The Expertise Check at kyc.co/expertise-check operationalises the SDD methodology. It scores an expert against the specific technical domain at issue in a dispute — producing a competence distance assessment at the moment of evaluation, not after the expert has already been heard.
Bad faith in legal proceedings is typically treated as an intentional state: a party who conceals, misrepresents, or strategically withholds is assumed to have chosen to do so. The existing legal framework responds by attempting to infer intent from behaviour — a slow, contested, and frequently inconclusive process.
This paper argues that procedural law itself can manufacture the structural conditions for bad faith behaviour without requiring deliberate intent from any individual actor. Specifically, where disclosure obligations are under-specified and silence is explicitly or implicitly protected by the system — through presumption rules, burden-of-proof allocations, or the absence of mandatory Variable B disclosure — the rational dominant strategy for a Passive Party in possession of adverse information is silence.
The paper applies mechanism design to legal procedure. A mechanism that makes silence the dominant strategy for an informed party is a mechanism that manufactures bad faith outcomes regardless of the subjective state of the parties. The Bad Faith Manufacturing Score (BFM) — a composite scored 0–100 — operationalises this observation as a measurable diagnostic derived from the structural features of the case: the nature of Variable B, the silence protection afforded by the applicable law, the asymmetry of information access, and the costs imposed on the Active Party by the Passive Party's rational non-disclosure.
The Bad Faith Check at kyc.co/bad-faith-check operationalises the BFM methodology across four analytical phases: fact mapping, structural bad faith scoring, forfeiture argument, and Loaded Gun Analysis (the prospective liability exposure created by the Passive Party's concealment — see Paper 04).
Paper 03 established the mechanism by which procedural silence manufactures bad faith outcomes. This paper identifies what happens next: the moment the Active Party proceeds — or does not proceed — the concealed Variable B detonates a second harm.
The paper proposes and examines three structural features that are present, in different legal domains, wherever a Passive Party's concealment transfers prospective liability onto an Active Party through the ordinary operation of law.
The Loaded Gun Law is a liability provision that fires on harm without ever examining whether the harm originates in a pre-existing condition. The law assigns responsibility by activity. It does not ask whether the affected asset, system, or counterparty was already compromised before the Active Party's act began. In construction: the party who touches the wall bears liability when the wall falls, regardless of its pre-existing structural state. The same pattern appears in product safety, data protection, medical device regulation, environmental liability, and employment law.
The Fiction of Uniformity is the implicit assumption the Loaded Gun Law makes: that the affected asset was in sound, uncompromised condition before the Active Party acted. The law never examines this assumption. It is a fiction embedded in the liability mechanism itself.
The Gridlock Paradox is the double-bind that results. If the Active Party proceeds: the Loaded Gun Law fires when Variable B causes harm, exposing the Active Party to liability they cannot escape. If the Active Party does not proceed: they bear the direct cost of inaction — blocked development, financing refused, litigation burn, lost income. Both paths impose material costs on the Active Party as a direct consequence of the Passive Party's concealment. The Passive Party cannot lose regardless of which path the Active Party chooses.
The Silence Multiplier captures how the Loaded Gun Law amplifies the value of the Passive Party's silence beyond mere non-disclosure: the concealment does not merely avoid a disclosure cost — it pre-loads the entire prospective liability of the Loaded Gun Law onto the Active Party as a contingent financial exposure that grows with every step the Active Party takes.
Phase IV of the Bad Faith Check at kyc.co/bad-faith-check — the Loaded Gun Analysis — operationalises this paper. It identifies the specific Loaded Gun Law applicable to any domain (construction, product safety, medical devices, data protection, environmental, employment), analyses whether it operates on a Fiction of Uniformity, states the full Gridlock Paradox double-bind, confirms whether both arms impose material costs, and quantifies the Silence Multiplier dynamic. The analysis is domain-agnostic: the tool finds whichever liability provision in whichever jurisdiction assigns responsibility by activity without examining Variable B.
These five papers are part of a larger body of work documented at moral.money/legal-innovation/ — The Procedural Gap Project. Four structural gaps in legal procedure. Five academic papers. Four diagnostic tools. One engineering constraint: closing loopholes that benefit bad-faith actors while ensuring the fixes are so tightly engineered they cannot be exploited in reverse.
Because what follows is dense, it is worth stating the whole of it once, plainly, before the apparatus arrives — so that the reader has the shape of the argument in hand throughout, and can see, at each step, which part of the problem that step addresses.
Here is what a civil dispute looks like from the outside. Two parties enter the process, and by the time they have done so, at least one of them is misrepresenting something. Not probably — as a matter of structure. Nobody litigates over a fact that both sides agree on. The act of formal dispute is a signal, before a single document is read, that at least one version of events on the record is false.
Both parties know this. And so both parties claim to be the victim. Both will be loud; both will have documentation; both will have professionals constructing the most coherent story available from the facts in their possession. From the outside, an honest party in a weak position and a dishonest party with a strong story can be indistinguishable. You cannot detect the misrepresentation by listening to who is more insistent, or by counting documents, or by measuring the confidence of the lawyers. The signal is not in what either party is asserting. It is in what one of them is not saying.
Every dishonest account has a gap in it. A document that should exist in an honest version of events does not. A period of time is described with less specificity than the periods around it. A question that any truthful answer to the dispute would naturally address goes, filing after filing, unanswered. The gap is not random — it sits at exactly the point where the honest version of events would require the party to say something that damages their own position. And that gap accumulates: the more the party writes, the wider the space grows between what they have said and what they have not said.
The existing system cannot see this gap. A court processes one document at a time. Each filing is read by professionals who were not present for the last one. Judges and lawyers change across proceedings that run for years. Nobody holds the whole record as a coherent narrative and compares what was asserted in an early filing against what was implied, avoided, or flatly contradicted in a later one. The record grows. The gap grows with it. And the gap is only visible to someone reading everything, in sequence, with no stake in what they find.
This is the Self-Reporting Fallacy — the paper's central observation. Every actor positioned to read a legal file has a specific stake in the answer it gives. A client trusts their own lawyer's account. A lawyer who filed a document is committed to its coherence. An insurer receives a summary of the file, not the file. A court sees what both sides choose to argue, not what either side has chosen not to raise. The result is a system capable of prosecuting the provable lie, and incapable — by design — of detecting the more common case: the assertion nobody checked, made by someone with every reason not to check it, sitting unread in a record that would have exposed it the moment someone compared what was said against what was not.
This paper specifies what that comparison requires. Three signals, taken together, make the gap detectable without requiring any inference about intent: how far the professional behind the filing sits from the matter's actual technical domain; what each party's documented circumstances make it rational for them to assert, withhold, or delay; and whether a filing's own positive factual claims survive contact with its own attachments. When they do not — when a party's filed evidence contradicts their own asserted position — that contradiction is the PI liability switch: the point at which a dispute between two parties becomes an independent professional-liability exposure that exists regardless of how the underlying case eventually resolves.
The lawyer in this dispute is not a player. They are the coach. They know more than the referee does. They have been briefed by their client on what actually happened, and their professional job is to build the best possible story from the facts available — within the rules, and ethically. That is not bullshit. That is advocacy.
But the coach carries a specific structural risk. If the story doesn't hold, both sides will look for someone to blame. The client who did not win will blame the coach for the loss. The opposing party — who may have added the firm as a second defendant the moment the contradiction surfaced — is already pointing. The coach has been in the room the whole time. They know things the referee does not. If, in advancing their client's position, they filed an assertion they knew was contested by their own file, or omitted a document they knew was damaging, or counselled their client on what to say and what to leave out: at the moment the story collapses, those actions become the new dispute.
The professional-indemnity insurer covering that coach is the one actor around the whole proceeding with both a clear financial stake in detecting this moment and, under current practice, no instrument for seeing it arrive. The PI insurer receives a notification that a claim has been made. They do not receive a reading of the file that shows when the professional's own evidence began to contradict the professional's own position. This paper builds that reading. Everything in the sections that follow is an attempt to do that one thing — locate the moment the gap became undeniable — without requiring any actor to have committed a provable lie.
Harry Frankfurt drew a distinction that the law has never formally used: a liar knows the truth and asserts its opposite, while a bullshitter is indifferent to the truth altogether, producing assertion calibrated to what will work rather than to what is so. The distinction matters because the two are almost impossible to tell apart from their output alone, and because the law's machinery for detecting deception — intent, knowledge, scienter — is built entirely for the first category and has essentially nothing to say about the second. A professional who files a claim their own evidence already contradicts may have lied. They may equally have never checked, because checking was not necessary to produce a filing that does its job. The document looks the same either way.
This paper argues that civil litigation, professional liability, and the institutions that sit above both have organised themselves around exactly this blind spot, and names the structural reason it persists: the Self-Reporting Fallacy. The fallacy is this: every actor positioned to read a legal file has a stake in the answer it gives. A client depends on trusting their own lawyer's account. A lawyer who filed a document is committed to its coherence. An insurer receives a summary of the file, not the file. A court sees what each side chooses to argue, not what either side has chosen not to mention. The result is a system that can prosecute the rare, provable lie but has no mechanism at all for the far more common case — the assertion nobody checked, made by someone with every reason not to check it, sitting undetected in a record that would have shown the gap to anyone reading without a stake in the outcome.
This paper specifies what a reading position with no stake in the outcome would need to do differently. It brings three signals together that the existing literature has only ever developed in isolation: how far the professional standing behind a filing sits from the matter's actual technical demands; what each party's documented circumstances make it rational for them to assert, withhold, or delay; and whether a filing's own positive factual claims survive contact with its own attachments. Demonstrated against one recurring event — the moment a professional's own filed evidence contradicts their own asserted position, converting a routine dispute into an independent professional-liability exposure that frequently dwarfs the underlying claim — the combination shows what becomes visible only once Frankfurt's distinction is taken as seriously in litigation as it has been in moral philosophy for forty years.
The paper makes four contributions. First, it names the Self-Reporting Fallacy and locates its cause precisely: not carelessness, but a predictable consequence of every actor around a dispute occupying a position that depends, in some specific and identifiable way, on the record's contradictions remaining unexamined. Second, it brings Frankfurt's distinction into a setting it has never been formally applied to. Third, it specifies the three-signal reading position. Fourth, it demonstrates the combination against one recurring event — the precise moment a professional's cover runs out — and argues that this event, long recognised by practitioners only in hindsight, becomes locatable at the time it occurs once a file is read this way.
The paper closes by locating the professional-indemnity insurance industry as the natural first market for instruments built on this method: the PI insurer is the one actor around litigation with both a financial stake in detecting the divergence event and, under current practice, no visibility into when it has occurred.
The Dispute Engine at kyc.co/dispute-engine is the prototype instrument referenced in the paper. It operationalises all three signals across an accumulating case record, running the internal documentary consistency test (Phase 8) against every filing and flagging the PI liability switch moment when detected.